Abstract: Relational data acquisition is an important field.In order to improve the utility and pertinence to data mining，this paper proposes a method for alternately constructing tuple-concepts and attribute-value-concepts from a relation，which is based on the hierarchies of attribute domains.These hierarchies of attribute domains are forbidden in the ordinary relations.This paper further defines rough-valued relations.Consequently，each tuple-concept is corresponding to a concept extent of some rough-valued relation，and it is proved that the concept lattice of the horizontal union of a relation and its rough-valued relations are exactly equal to the horizontal union of concept lattices of these factor relations.Thus，one can construct a complex concept lattice from simple ones，which may reduce complexity of constructing concept lattice of a complex relation.

Key words: rough-valued relations, concept lattices, tuple-concepts, attribute-value-concepts, horizontal union

摘要： 关系数据获取是一个很重要的研究领域。为了提高挖掘的针对性和有效性，提出了一种基于属性取值域层次划分的元组概念与属性值概念的交替获取方法。这些属性取值域上的划分是关系中所蕴涵的重要信息。另外，从属性取值域的层次划分出发，定义了一类扩展的关系-粗值关系。每个元组概念都是某个粗值关系中的概念外延。因此，元组概念获取可转化为构造关系的概念格，并证明一个关系与某些粗值关系的横向并的概念格等于它们各自概念格的横向并。由简单概念格的并构造复杂概念格可减少构造概念格的复杂性。

关键词: 粗值关系, 概念格, 元组概念, 属性值概念, 横向并